Ambient Formaldehyde Measurements Made at A

Home , Mace (spray)

Atmos. Chem. Phys., 6, 2711–2726, 2006 www.atmos-chem-phys.net/6/2711/2006/ Atmospheric © Author(s) 2006. This work is licensed Chemistry under a Creative Commons License. and Physics

Ambient formaldehyde measurements made at a remote marine boundary layer site during the NAMBLEX campaign – a comparison of data from chromatographic and modiﬁed Hantzsch techniques

T. J. Still1, S. Al-Haider1, P. W. Seakins1, R. Sommariva1, J. C. Stanton1, G. Mills2, and S. A. Penkett2 1School of Chemistry, University of Leeds, Leeds, LS2 9JT, UK 2School of Environmental Sciences, University of East Anglia, Norwich, NR4 7TJ, UK Received: 9 August 2005 – Published in Atmos. Chem. Phys. Discuss.: 6 December 2005 Revised: 13 June 2006 – Accepted: 13 June 2006 – Published: 6 July 2006

Abstract. Ambient formaldehyde concentrations are re- that oxygenated VOCs such as ethanal and methanol are very ported from the North Atlantic Marine Boundary Layer Ex- significant sources of formaldehyde in the air masses reach- periment (NAMBLEX) campaign at Mace Head on the west ing Mace Head. coast of Eire during August 2002. The results from two techniques, using direct determination via gas chromatography and the Hantzsch technique, show similar trends but a significant off set in concentrations. For westerly air flows 1 Introduction characteristic of the marine boundary layer, formaldehyde concentrations from the gas chromatographic and Hantzsch technique ranged from 0.78–1.15 ppb and 0.13–0.43 ppb, re- The North Atlantic Boundary Layer Experiment, NAM- spectively. Possible reasons for the discrepancy have been BLEX, took place at Mace Head, Eire, during July and Au- investigated and are discussed, however, no satisfactory ex- gust 2002 to help quantify our understanding of photochem- planation has yet been found. In a subsequent laboratory in- ical oxidation processes in clean and moderately polluted tercomparison the two techniques were in good agreement. environments. Objectives included quantifying the role of The observed concentrations have been compared with halogen species in the marine boundary layer (MBL), study- previous formaldehyde measurements in the North Atlantic ing the reactive nitrogen budget and formation of new par- marine boundary layer and with other measurements from ticles. Objectives particularly relevant for our formaldehyde the NAMBLEX campaign. The measurements from the (HCHO), measurements included model/measurement com- Hantzsch technique and the GC results lie at the lower and parisons for radical species and the role of reactive hydro- upper ends respectively of previous measurements. In con- carbons in the MBL. Formaldehyde plays an important role trast to some previous measurements, both techniques show in determining radical concentrations, influencing both HOx distinct diurnal profiles with day maxima and with an ampli- formation and removal. Due to its importance and difficulties tude of approximately 0.15 ppb. Strong correlations were ob- in determining accurate concentrations (Gilpin et al., 1997) served with ethanal concentrations measured during NAM- formaldehyde concentrations were measured using two dif- BLEX and the ratio of ethanal to formaldehyde determined ferent techniques. The University of Leeds (UoL) used a by the gas chromatographic technique is in good agreement chromatographic method based on the detection of separated with previous measurements. HCHO with a helium ionization detector (Hunter et al., 1998, Some simple box modelling has been undertaken to inves- 1999; Hopkins et al., 2003). The University of East Anglia, tigate possible sources of formaldehyde. Such models are UEA, used a version of the Hantzsch reaction (Cardenas et not able to predict absolute formaldehyde concentrations as al., 2000) where ambient HCHO was scrubbed into solutions they do not include transport processes, but the results show and derivitized by reaction with 2,4 pentadione and ammonia. The adduct was then detected via fluorescence at 510 nm Correspondence to: P. W. Seakins following UV excitation by a mercury lamp. The two sys- ([email protected]) tems were located at sites approximately 200 m apart.

Published by Copernicus GmbH on behalf of the European Geosciences Union. 2712 T. J. Still et al.: Ambient HCHO at NAMBLEX

Table 1. Some previous measurements of formaldehyde in the marine environment.

Reference Campaign and Date Location Platform Technique [HCHO] in Notes clean air/pptv Harris et al. (1992) “Polarstern” 1988 North Atlantic Ship TDLS 650 Measurement is mean value over 40–45◦ N. No diurnal variation Tanner et al. (1996) NARE 1993 Nova Scotia Coastal site DNPH 200–400 Summer measurement Solberg et al. (1996) EMEP 1994-5 Mace Head Coastal site DNPH 200–500 Winter measurement. Measurements at other sites show summer maxima. Cardenas et al. (2000) ACSOE 1996 Mace Head Coastal site Hantzsch 200–450 Summer measurement. Weller et al. (2000) ALBATROSS 1996 North Atlantic Ship Hantzsch 400–500 No diurnal variation observed. Fried et al. (2002) NARE 1997 Newfoundland Plane TDLS 410±150 Wagner et al. (2001) INDOEX 1999 Indian Ocean Ship TDLS 430±100 Diurnal variation with amplitude of ∼200 pptv observed.

In the very remote environment, methane is the major In a modelling study of formaldehyde production in the re- source of formaldehyde via the reactions: mote Indian Ocean, Wagner et al. (2002) showed that in the presence of only 2 pptv of NO approximately 50% of CH3O2 O2 OH + CH4 −→ H2O + CH3O2 (R1) reacts via Reaction (R9). There are two major chemical removal processes for CH3O2 + HO2 → CH3OOH + O2 (R2) formaldehyde. Firstly, reaction with OH (Reaction R10) which, via the subsequent rapid reaction of HCO with oxy- CH O + CH O → 2CH O + O (R3) 3 2 3 2 3 2 gen (Reaction R11), is a major route for conversion of OH to HO . CH3O2 + CH3O2 → HCHO + CH3OH + O2 (R4) 2 OH + HCHO → H O + HCO (R10) CH OOH + hν → CH O + OH (R5) 2 3 3 −8 HCO + O2 → HO2 + CO(τ=4 × 10 s) (R11) OH + CH OOH → HCHO + H O + OH (R6) 3 2 Secondly, photolysis, which can act as a signiﬁcant free rad-

OH + CH3OOH → CH3O2 + H2O (R7) ical source. The efﬁciency of HCHO as a radical source de- pends on the branching ratio between the molecular and rad- CH3O + O2 → HO2 + HCHO (R8) ical channels (Pope et al., 2005): + → + Under such low NOx conditions, the methyl peroxy radical HCHO hν H2 CO (R12) (CH3O2) formed from the reaction of OH with CH4 pre- HCHO + hν → H + HCO (R13) dominantly reacts with other peroxy radicals. Reaction with Formaldehyde, and carbonyls in general, have been shown HO2 leads to CH3OOH which effectively acts as a reservoir species on route to HCHO formation (although it can to be major sources of HOx radicals in the urban winter at- be rained out). Self reaction leads more directly to HCHO mosphere (Heard et al., 2004), but formaldehyde is also an formation. Reaction (R8) is very rapid and other reactions important HO2 source in the remote free troposphere. Frost of methoxy radicals do not need to be considered. Addi- et al. noted that the importance of formaldehyde as a radi- tional sources of formaldehyde include higher hydrocarbons cal source will increase at higher, increasingly drier altitudes and oxygenated VOCs such as methanol or acetaldehyde. (Frost et al., 2002). Table 1 lists some previous formaldehyde measurements In the presence of sufﬁcient NO, CH3O2 radicals react directly with NO to form methoxy radicals and subsequently in remote MBL environments, focusing particularly on stud- formaldehyde via Reaction (R8). ies in the North Atlantic environment. Measurements have been with a variety of techniques and from airborne, ship- CH3O2 + NO → CH3O + NO2 (R9) borne and coastal platforms. Average values range from

Atmos. Chem. Phys., 6, 2711–2726, 2006 www.atmos-chem-phys.net/6/2711/2006/ T. J. Still et al.: Ambient HCHO at NAMBLEX 2713

∼200–1000 pptv. A seasonal dependence has been observed 2.2 UoL apparatus for formaldehyde measurements in the EMEP programme (Solberg et al., 1996) peaking during the summer months. The UoL instrument used during the NAMBLEX campaign There is conflicting evidence on meridional variations with was based on a gas chromatographic (GC) system as de- measurements in the North Atlantic showing both positive scribed by Hopkins et al. (2003). The sampling inlet was and negative variations with increasing latitude (Harris et al., placed 2 m above the ground and consisted of ∼12 m 1/400 1992; Weller et al., 2000). Fried and co-workers have carried PFA tubing, and was pumped at a speed of 1 slm. The sam- out several airborne campaigns, their measurements show a ple passed through a 6.4 ml sample loop (Silco Steel). During general decrease in formaldehyde concentrations with alti- injection, helium carrier gas (CP grade, BOC, 9 ml min−1, tude (Fried et al., 2002, 2003). backing pressure 50 psi, further purified via passing through As can be seen from Table 1, a variety of experimental liquid nitrogen traps) was diverted through the loop, sweep- techniques have been deployed. Partially, this reflects the ing the sample onto the column (50 m, 0.32 mm id, 100% difficulties in making reliable measurements on this impor- dimethyl polysiloxane, WCOT column, 5 µm phase thick- tant atmospheric intermediate. The deployment of a vari- ness, CP-Sil 5CB Chrompack, Netherlands) and refocused at ety of techniques has highlighted potential systematic errors the head of the column in a liquid nitrogen trap. Following in the measurement techniques. A good review of several the elution of the untrapped air, the analytes were released techniques and the results of a field intercomparison can be and were separated in the column and detected using an ar- found in the work by Gilpin et al. (1997) and very recently gon doped (1% Ar in He, BOC, CP grade), pulsed discharge the results of an HCHO field (urban) intercomparison from helium ionisation detector (pdHID) (Model D4, VICI AG, the FORMAT programme have been published (Hak et al., Schenkon, Switzerland). The detector flow was 30 ml min−1 2005). maintained with a calibrated restrictor. After the elution of formaldehyde the column flow was reversed and the column was back flushed (30 ml min−1, 70 s, backing pressure 2 Experimental 60 psi) to prevent water reaching the detector or the build-up of heavier weight material on the column. The sample cycle 2.1 Site time throughout the campaign was 5.5 min. The detection limit during the campaign was 42 pptv; this NAMBLEX took place at the Mace Head observatory on the was calculated taking the minimum detectable peak to have ◦ 0 00 ◦ 0 00 west coast of Eire (53 19 34 N, 9 54 14 W) during July a signal to noise ratio of 3:1. The system used a formalde- and August 2002. A majority of formaldehyde measure- hyde gas phase standards generator for calibration (KIN- ments were taken between 1–21 August 2002. The location TEK, LaMarque, TX). A permeation tube containing poly- of the site is shown in Fig. 1. Air arriving between the angles meric formaldehyde with a known emission rate at a set refer- ◦ of 180 and 300 is free from any local land influence. ence temperature (333 K) was held in a stabilized oven (tem- Five day back trajectories were calculated based on the perature variation <±0.1 K) flushed with a constant flow rate wind field analysis produced by the European Centre of (10.0 sccm) of UHP nitrogen. Delivery concentrations of 4– Medium Range Weather Forecasts (ECMWF) and were used 50 ppbv (during campaign), or subsequently 2–50 ppbv (with to classify the origins of the air masses arriving at Mace increased flow rate), were obtained by diluting the oven flow Head. During the campaign local meteorological measure- with additional known flow rates of UHP nitrogen. The de- ments were made by a wind profiler and its observations livery tube was connected to the GC sample inlet via a “tee” provided a record of the boundary layer structure, including junction with a positive flow out of the vent, ensuring that the measurements of average wind speed and direction (Norton delivery pressure was at one atmosphere. The precision was et al., 2006). The local measurements suggest that for west- attained from the standard deviation of replicate calibration erly winds, the ECMWF trajectories and local measurements factors and was calculated to be 3%. The uncertainty was at- are in good agreement. For north-easterly wind flows, char- tained from the errors associated with the standards generator acteristic of the early part of the campaign, local sea breezes and was determined to be 9% (2 σ ). dominated. However, the chemical signatures of typical anthropogenic pollutants such as CO and C2H2 suggest that the 2.3 UEA apparatus site is still receiving air broadly characteristic of origin of the air mass, albeit potentially moderated by local meteorology. The fluorescence technique is based on the Hantzsch reac- The inset to Fig. 1 shows the location of the two formalde- tion which is a liquid phase reaction of formaldehyde fol- hyde measurements. The UEA apparatus was co-located lowed by fluorescence detection of the resulting adduct. This with a majority of the other instrumentation ∼100 m distant technique requires formaldehyde to be transferred from the and at ∼10 m elevation from the average high tide mark. The gas phase into liquid phase, achieved via a stripping solution UoL instrumentation was located a further ∼200 m inshore of 0.1 N H2SO4 (made up from ACS reagent grade H2SO4, and elevated by a further ∼20 m. Aldrich) at room temperature, which is pumped through a www.atmos-chem-phys.net/6/2711/2006/ Atmos. Chem. Phys., 6, 2711–2726, 2006 2714 T. J. Still et al.: Ambient HCHO at NAMBLEX

Fig. 1. Location of the Mace Head observatory and the relative positioning of the two formaldehyde instruments. coil (45 cm long, 10 turns of 3 mm o.d. Pyrex) and forced into tion uncertainties and stripping efficiency uncertainties) were contact with gaseous formaldehyde. Ambient formaldehyde ±6.3% and ±13.2%, respectively with a minimum uncer- was sampled at a height of ∼5 m through approximately 10 m tainty of ±25 pptv. The detection limit (three sigma of zero of 3/800 PFA tubing. The pumping speed was a total of 15 slm noise) of the instrument varied between 38 and 69 pptv dur- giving a very short residence time in the sampling line, and ing the campaign. Zeros were performed every 5 h by scrub- only small decrease in pressure in the sample line. The sam- bing the sample with a charcoal filter. ple was introduced into the instrument at 1.5 slm through a 1 m length of 1/400 PFA tubing. 2.4 Intercomparisons and artefacts The Hantzsch process is based on the liquid phase reaction of formaldehyde with acetylacetone (2,4 pentadione) During the campaign the calibration sources for both instru- and ammonia to produce diacetyldihydrolutidine, (DDL) that ments were exchanged and compared. When used on the is excited at 412 nm (Hg-Phosphor 215 lamp, Jelight, USA) UoL instrument, the UEA calibration source produced val- and the fluorescence is detected at 510 nm with a photomul- ues within 2% of those generated using the UoL source. It tiplier tube (Hamamatsu). was therefore concluded that the calibration sources were in The detection limit for this instrument was in the region good agreement for the campaign period. of 50 pptv (Cardenas et al., 2000). The instrument was also The two instruments also took part in a laboratory inter- calibrated using a permeation source (KIN-TEK LaMarque, comparison at the National Physical Laboratories (Tedding- TX). The calibrant introduced into the instrument via a “Tee- ton, UK) during June 2003. In these experiments all the piece” to vent the excess and ensure the calibration was de- instruments sampled from a common chamber (NPL Stan- livered at atmospheric pressure. The mean precision and dard Atmosphere Generator). HCHO concentrations in the accuracy of the instrument (based on zero noise, calibra- chamber were calculated from permeation rates of a known

Atmos. Chem. Phys., 6, 2711–2726, 2006 www.atmos-chem-phys.net/6/2711/2006/ T. J. Still et al.: Ambient HCHO at NAMBLEX 2715

Fig. 2. Ozone artefact for (N) UoL apparatus, () UEA apparatus as determined during experiments at NPL, 2003. standard and measured dilution flows. All the instruments calibration processes). Secondly there is a reasonable cor- performed linearity and stability checks and a blind sam- relation between the two measurements, and in some cases, pling. Various concentrations of ozone were generated and for example that shown in the inset to Fig. 3, the pattern of introduced into the chamber to investigate possible artefacts. variation is almost identical. Both the instruments responded well to the linearity (2– Figure 4 shows a correlation plot of the results. There 8 ppbv) and stability experiments (apparatus run overnight at is significant scatter in the data which may be associated 0.5 ppbv) and reported results within experimental errors for with sampling slightly different air masses either due to lo- the blind testing, although the concentration, at ∼7 ppbv, was cal meteorology or to differences in the sampling techniques. significantly higher than ambient concentrations in the MBL. The UoL apparatus is essentially a grab sample taken every Both instruments showed a positive interference for ozone as 5.5 min, whereas UEA data are based on averaging over a shown in Fig. 2. one minute period. The difference in the UoL and UEA data The magnitude of the ozone artefact on the UoL instru- appears to be in an offset rather than in the gradient of the ment was significantly greater, but was reproducible, both correlation plot. at NPL and during subsequent experiments at Leeds. The Given the good agreement on the trends in formaldehyde results reported below for both instruments have been cor- concentrations, possible reasons the systematic error include; rected for ozone measurements as recorded at Mace Head by sample losses in the lines or scrubbing processes, calibration the FAGE group using a UV photometric analyser. errors, interferences from other gases and significant blanks in the UoL apparatus. The University of Leeds measurements showed no varia- 3 Results tion with the length of sampling tube in either field or laboratory tests. Given that the same material was used for sam- 3.1 Comparison of time series from UoL and UEA mea- pling in the UEA apparatus, we would not expect any signif- surements icant losses in that system either. Significant loss processes in the UEA instrument would be very surprising since the Figure 3 shows the formaldehyde time series for both instru- whole wetted path of the sample is PFA and no loss processes ments. Two observations are immediately apparent. Firstly, on PFA have been seen in numerous laboratory tests. The the results from the UoL are significantly higher, and out- difference in pressure between the sample drawn down the side the combined random errors of both techniques (9% for 3/800 PFA and the calibrant delivered directly into the instru- UoL and 13% for UEA based on errors associated with the ment is at most 10%. Both losses and differences in pressure www.atmos-chem-phys.net/6/2711/2006/ Atmos. Chem. Phys., 6, 2711–2726, 2006 2716 T. J. Still et al.: Ambient HCHO at NAMBLEX

Fig. 3. Time series of formaldehyde concentrations from GC and ﬂuorescence techniques. All data corrected for ozone interferences.

Fig. 4. A graph showing the weighted bivariant regression between the GC and the ﬂuorescence technique for all data (solid line), and the 1:1 line (dotted line). would lead to an underestimation of HCHO, but would be In general, calibration errors do not appear to be a problem expected to show a difference in slope rather than an offset as the calibration systems were compared during the cam- in the scatter plot. paign. The UEA system requires transfer from the gas to

Atmos. Chem. Phys., 6, 2711–2726, 2006 www.atmos-chem-phys.net/6/2711/2006/ T. J. Still et al.: Ambient HCHO at NAMBLEX 2717

Table 2. Summary of UoL and UEA formaldehyde concentrations for the various air masses encountered during NAMBLEX.

UoL UEA Air Mass Origin 24 h average Range/ppbv 24 h average Range/ppbv [HCHO]/ppbv [HCHO]/ppbv Continental 1.20±0.13 0.98–1.58 0.37±0.11 0.13–0.73 Westerly 0.96±0.09 0.78–1.15 0.27±0.07 0.13–0.43 Anticyclonic 0.92±0.15 0.72–1.31

liquid phase. This process is always highly efficient, but is ing from south to north westerly). It is possible to further slightly temperature dependent (92–98% as temperature de- subdivide this period into air masses with NW, W, and SW creases from 298–278 K). The temperature of the laboratory origins, however, analysis showed no significant variation fluctuated by ±5 K and variations in sampling versus calibra- in formaldehyde concentrations with these finer components tion temperature are an additional small source of error. and hence all westerlies are considered together. Over the As described in the previous section ozone artefacts were last few days of the formaldehyde campaign the wind was detected in both systems and the reported data have been suit- still westerly but from an anticyclonic weather system with ably corrected. Corrections were typically of the order of low wind speeds. Summaries of the formaldehyde concen- 300–400 pptv formaldehyde. One other possible artefact par- trations for each period can be found in Table 2. ticularly relevant to the MBL is the presence of water vapour. Tests performed at NPL were inconclusive and further work Not surprisingly, formaldehyde concentrations were high- is planned. It should be noted that there was no systematic est during the initial continental air flow. Concentrations variation in the offset with relative humidity. Blanks were of hydrocarbons and CO are also elevated during this pe- recorded on a daily basis for the UoL apparatus. These were riod. Thereafter, concentrations decreased and a diurnal obtained by attaching a cylinder of zero nitrogen (Premier pattern became more apparent, Fig. 5a, with concentrations Grade, Air Products) via 1/400 Teflon tubing to the inlet port during the westerly air flows varying from an early morn- of the GC. A “tee” junction was placed upstream of the sam- ing (03:00–05:30 GMT) minimum of ∼850 pptv to a peak ple inlet to restrict the upstream pressure to 1 atmosphere. value of ∼1050 pptv between 14:00–16:00 GMT (local time The zero air was drawn into the GC via the pump in the is GMT+1). During the latter part of the anticyclonic west- normal way. An ozone artefact for the UoL apparatus has erly airflows, the amplitude of the diurnal profile appeared to already been described. It should be noted that blank mea- increase. There were also concurrent increases in acetylene surements were made with dry gas, our planned work on hu- concentrations suggesting some local contamination during midity effects may determine whether this is an important these periods. issue. We cannot rule out the possibility of other undetected artefacts. Obviously one, or both, instruments were subject to systematic errors during NAMBLEX, but either these had been 3.3 Average values and diurnal profiles from UEA rectified by the following summer, or are only apparent during real air, rather than laboratory sampling. We are therefore at a loss to explain the differences in measured concentrations at Mace Head during NAMBLEX and in the following Results from the fluorescence apparatus show generally sim- section the results of each apparatus are presented separately. ilar behaviour. Concentrations are elevated at the start of The measurements are compared with other measurements the campaign, but appear to remain constant between 200– and various models in the discussion section. 500 pptv from the 3–14 August. As with the UoL data for the westerly winds, a pronounced diurnal variation is observed, 3.2 Average values and diurnal profiles for UoL apparatus as shown in Fig. 5b, with an amplitude similar to that observed by the UoL. The campaign can broadly be divided into three different air masses. From 1–5 August the site experience continen- The concentrations measured on 15 August are signifi- tal air masses which had originated in Scandinavia and had cantly lower than other periods during the westerly airflow passed over northern England. This was followed by a long with values between 50–100 pptv. After the 16 August data period (6–17 August) of relatively long trajectories (high from the UEA apparatus became patchy due to intermittent wind speeds) with a significant westerly component (rang- lamp failure. www.atmos-chem-phys.net/6/2711/2006/ Atmos. Chem. Phys., 6, 2711–2726, 2006 2718 T. J. Still et al.: Ambient HCHO at NAMBLEX

Fig. 5a. Diurnal variation of formaldehyde concentrations as measured by UoL during westerly airﬂows (6–17 August 2002). Error bars represent one standard deviation of concentrations at that particular timestamp.

Fig. 5b. Diurnal variation of formaldehyde concentrations as measured by UEA during westerly airﬂows (6–16 August 2002). Error bars represent one standard deviation of concentrations at that particular timestamp.

4 Discussion Formaldehyde measurements at Mace Head are relatively limited. A measurement campaign was conducted at Mace 4.1 Comparison with previous marine boundary layer mea- Head under the EMEP programme, with formaldehyde be- surements ing monitored from November to April using 2,4 DNPH car- tridges with subsequent off-line analysis by HPLC (Solberg et al., 1996). Samples were collected twice weekly, each Some results from previous campaigns are presented in Ta- measurement consisting of 8 h sampling centred around mid- ble 1. We have focused predominantly on measurements day, with subsequent off line analysis at NILU. Solberg et from the North Atlantic although some other campaigns are al. reported monthly averages between 200–500 pptv over included.

Atmos. Chem. Phys., 6, 2711–2726, 2006 www.atmos-chem-phys.net/6/2711/2006/ T. J. Still et al.: Ambient HCHO at NAMBLEX 2719 this winter period. Direct comparison with our summer agreement. Harris et al. report a mean formaldehyde concen- data cannot be made, however, the winter concentrations tration between 40–45◦ N of 650 pptv, with concentrations appear to correlate well with those made at a Norwegian decreasing slightly during the cruise south. site (Birkenes), where summer concentrations reach almost When used to compare with formaldehyde concentrations 1 ppbv. Solberg et al. report annual variations in formalde- at Mace Head, the results of other campaigns need to be in- hyde levels from a number of rural sites around Europe and terpreted carefully. For air and shipborne measurements vari- all show summer maxima. ation with latitude will be strongly dependent on the origins Formaldehyde was measured at Mace Head during the of the air mass reaching the receptor at that point and may ACSOE 96 and 97 campaigns. A majority of the measure- not be typical of longer term measurements at that latitude. ments were made using an earlier version of the fluoromet- Both Fried et al. (2002) and Weller et al. (2000) report a de- ric technique, and a more limited data set was also recorded crease in average formaldehyde concentrations with latitude, using tuneable diode laser spectroscopy (TDLS) (Cardenas however, because the air masses trajectories and ocean cur- et al., 2000). Typical values for westerly air masses were rents vary across the region, this does not mean that the lat- 200–400 pptv from the fluorometric technique, whereas the itudinal variation will be the same over the eastern Atlantic. TDLS produced higher, but scattered values with a range of Finally, the EMEP data showed significant annual variation 0–1000 pptv. in formaldehyde concentrations (as would be expected from A number of formaldehyde measurements were made dur- a photochemical intermediate) and therefore results from dif- ing the North Atlantic Regional Experiment (NARE) cam- ferent seasons should be interpreted with care. paigns. During NARE 1993, Tanner et al. (1996) mea- The final campaign discussed in this section took place in sured a range of carbonyl compounds using derivatization via the Indian Ocean during spring 1999 (Wagner et al., 2001). 2,4 DNPH/HPLC at Cherbourg Point, Nova Scotia. Typical Given the location, comparison with Mace Head concen- values for clean air at 200–400 pptv are in good agreement trations are difficult, however, results from the INDOEX with the UEA data. Measurements were only made every six campaign, (where formaldehyde was measured by TDLS) hours, so no diurnal trends are reported. Tanner et al. report a are interesting for several reasons. Firstly, even for trajec- relationship between formaldehyde and ethanal with the lat- tories that had been over the open ocean for seven days, ter being approximately 50% of the formaldehyde concentra- and hence lost any continental formaldehyde, concentrations tions. This observation will be discussed further below. were still significant (∼500 pptv). A direct oceanic source is During NARE 1997 a number of airborne measurements not expected and therefore this is good evidence of signifi- were made from St John’s, Newfoundland, using both TDLS cant HCHO production from long-lived precursors, includ- and the coil/DNPH technique (Fried et al., 2002). Gener- ing oxygenated VOC. Secondly, a strong positive correlation ally the agreement between the two techniques was good was noted with CO for a variety of air masses including very with a gradient on the scatter plot of 0.98±0.07, however, aged air (HCHO/CO ≈3×10−3). Finally, during periods of approximately 30% of the data points lay outside the com- consistent air flow, diurnal variations in [HCHO] were ob- bined 2 σ uncertainties. Median values in clean background served with an amplitude of approximately 200 pptv in the ◦ air from 0–2 km, between 35–55 N are 400 pptv for TDLS northern hemisphere, consistent with the observed ampli- (range 300–900 pptv for lowest altitude measurements) and tudes of both UEA and UoL instruments during clean west- 410 pptv for the CDNPH (range 200–800 pptv for lowest al- erly airflows. titude measurements). Average concentrations decrease with both altitude and latitude. 4.2 Comparison with other NAMBLEX measurements Two shipborne campaigns have monitored formaldehyde in the North Atlantic. The ALBATROSS campaign took The NAMBLEX campaign was characterised by relatively place during October and November 1996 with the cruise polluted air during the first part of the campaign, with a pe- being a meridional track from 60◦ N to 45◦ S. Formalde- riod of consistent, strong, westerly, clean air and finally an- hyde was monitored via a commercial apparatus based on ticyclonic conditions, originating in the North Atlantic. An- the Hantzsch reaction (Weller et al., 2000). A variety of dif- thropogenic pollutants are therefore expected to be high dur- ferent air masses were encountered. Typical concentrations ing the initial part of the campaign, and thereafter relatively when the ship was between 20–40◦ N and intercepting trajec- low, although the low wind speeds under anticyclonic condi- tories similar to the westerlies and south- westerlies reach- tions may allow for influence by local sources. The acetylene ing Mace Head were in the region of 300–800 pptv. Harris and CO concentrations shown in Fig. 6 broadly match the ex- et al. (1992) report data from the 1988 “Polar Stern” cruise pected behaviour. which took place during September and October from 45◦ N If the correlation between formaldehyde and CO, observed into the southern hemisphere. Formaldehyde was predomi- during the INDOEX campaign by Wagner et al. (2001) ap- nantly measured using TDLS with a small number of com- plied under NAMBLEX conditions, then formaldehyde con- parative experiments obtained with DNPH sampling and sub- centrations during westerly airflows would be predicted to sequent analysis via HPLC. The two methods were in good be in the region of 200–300 pptv, in good agreement with the www.atmos-chem-phys.net/6/2711/2006/ Atmos. Chem. Phys., 6, 2711–2726, 2006 2720 T. J. Still et al.: Ambient HCHO at NAMBLEX

Fig. 6. Signatures of anthropogenic tracers. Acetylene and CO data were provided by the University of York and Mace Head research station, respectively.

UEA data. However, Fig. 6 shows a strong correlation be- from the UoL is good with both showing the same gradual tween CO and acetylene, a primary pollutant. Therefore CO decline in concentrations at the beginning of the campaign, concentrations at Mace Head may correlate more strongly followed by consistent and relatively low levels during the with primary sources rather than photochemical production periods of westerly air flow. and the relationship between formaldehyde and CO observed The correlation plot of acetaldehyde con- during highly aged air in the Indian Ocean may not apply. centrations with the UoL formaldehyde data Methane levels are likely to be slightly elevated in pol- ([HCHO]=(1.25±0.29)[CH3CHO]+0.46±0.19) shows luted air masses, but formaldehyde is also formed during the an intercept of 0.46 ppbv on the HCHO axis. An intercept atmospheric oxidation of virtually all higher hydrocarbons would be expected given that a significant fraction of and oxygenated VOCs (OVOC), and due to their higher pho- formaldehyde will be formed from methane or from frag- tochemical loss rates, these latter species are likely to be the mentation processes that by-pass acetaldehyde production. dominant influence on formaldehyde concentrations. A significant positive correlation between formalde- A good example of a formaldehyde precursor is acetaldehyde and acetaldehyde was observed by Tanner et hyde (ethanal). Both photolysis and reaction with OH can al. (1996) during measurements at a coastal site in lead to formaldehyde formation: Nova Scotia during NARE 1993. They reported

O2 [HCHO]ppbv=1.9[CH3CHO]ppbv+0.22 with an r coefﬁcient CH3CHO + hν −→ CH3O2 + CO + HO2 (R14) of 0.62. Errors on the gradient and intercept are not reported, O ± OH + CH CHO −→2 CH C(O)O + H O (R15) but from the scatter of the plot, are likely to be at least 33% 3 3 2 2 and therefore comparable with our parameters derived ob- + → + + CH3C(O)O2 RO2 CH3 CO2 products (R16) servations between acetaldehyde and the UoL formaldehyde Figure 7a shows a comparison of the time series for ac- data. Monthly averaged formaldehyde and acetaldehyde con- etaldehyde and formaldehyde, with Fig. 7b showing the cor- centrations were reported by Solberg et al. (1996) from a responding correlation plots with the UoL data. Acetalde- number of remote European sites including Mace Head. In hyde was measured by the University of York (Hopkins et all cases formaldehyde concentrations were greater than ac- al., 2003; Lewis et al., 2005) concentrating VOC and OVOC etaldehyde, typically about a factor of two during the summer from ∼1 l of air onto a solid trap followed by rapid thermal months. desorption, separation via GC with detection by FID. The The UEA formaldehyde concentrations are comparable in general correlation between the CH3CHO and H2CO data value with the measured acetaldehyde concentrations. Given

Atmos. Chem. Phys., 6, 2711–2726, 2006 www.atmos-chem-phys.net/6/2711/2006/ T. J. Still et al.: Ambient HCHO at NAMBLEX 2721

Fig. 7a. Time matched series of formaldehyde (UoL) and acetaldehyde (University of York).

Fig. 7b. Correlation plot of UoL formaldehyde vs. acetaldehyde. The regression equation is [HCHO]=1.25(±0.07)×[CH3CHO] +0.46(±0.03), where the errors are 1 σ. our current understanding of formaldehyde and acetaldehyde eral form the same type of radicals as OH initiation. Al- photo oxidation these observations appear to be incompati- though concentrations of Cl atoms are much lower than ble. OH in the MBL (typically 5×103 cm−3 vs. 1×106 cm−3), the magnitude of the abstraction rate coefﬁcients (typically Methanol is another signiﬁcant OVOC precursor for >5×10−11 cm3 molecule−1 s−1 (Qian et al., 2002; Seakins formaldehyde. Reaction with OH leads to HCHO via re- et al., 2004)) is such that Cl initiated chemistry can be signif- action of either CH3O isomer formed by the initial abstrac- icant. tion reaction. In the MBL reaction of Cl atoms with hydrocarbons and OVOCs may also need to be considered (Ramacher et al., 1999). Cl initiated reactions tend to be less selective in the position of abstraction, but in gen- www.atmos-chem-phys.net/6/2711/2006/ Atmos. Chem. Phys., 6, 2711–2726, 2006 2722 T. J. Still et al.: Ambient HCHO at NAMBLEX

Fig. 8a. Comparison of measured and modelled formaldehyde for 9–10 August 2002 (JD221–222).

Fig. 8b. Comparison of measured and modelled formaldehyde for 15–21 August 2002 (JD227–233).

4.3 Modelling however, zero dimensional box models give an indication of the expected concentrations from various chemical schemes The simplicity of the chemistry in the MBL allows for the and highlight the relative importance of various formation construction of relatively simple models to test our under- and removal channels. A good example of a box model is standing of photochemistry in this environment. The ideal described by Wagner et al. (2001, 2002) and used to interpret molecules for comparison are those with very short lifetimes, shipborne measurements in the remote Indian Ocean from so that transport can be ignored and realistic comparisons can the INDOEX campaign. In this case methane was the domi- be made with box models. Hydroxyl and hydroperoxy radi- nant source of formaldehyde and good agreement (typically cals are ideal examples and comparisons between model pre- within 20%) was found between measurements and modelled dictions and measurements have been made in a number of formaldehyde concentrations. However, as shown in Table 3, environments, including the marine boundary layer (Carslaw in other campaigns the models can either under or over pre- et al., 2002; Heard et al., 2004; Sommariva et al., 2004). The dict formaldehyde concentrations. lifetime of formaldehyde (typically ∼4 h under midday con- In this study we have used the Master Chemical Mech- ditions) is such that transport should really be considered, anism (MCM) (Saunders et al., 2003) to simulate the

Atmos. Chem. Phys., 6, 2711–2726, 2006 www.atmos-chem-phys.net/6/2711/2006/ T. J. Still et al.: Ambient HCHO at NAMBLEX 2723

Table 3. Summary of the results from the measured-modelled comparisons of formaldehyde from the literature. a Daily mean. b Value read from a graph (Peak diurnal value). c Result from summer intensive.

Authors Environment Model constraints Typical [HCHO]/pptv Under/Over prediction Weller et al. (2000) Marine, Atlantic Photochemical box model based 580±160a Under predicted by a ◦ ◦ 48 N–35 S on CH4 + CO photo oxidation. factor of 2 Ayers et al. (1997) Coastal, Cape Grim, Photochemical box model based 400±50b Under predicted by a Australia on CH4 + CO photo oxidation in- factor of 2 cludes dimethyl sulphide mechanism Wagner et al. (2002) Marine, Indian Ocean MCM 2. Photochemical oxida- 200±70a Over predicted by tion of CH4 and NMVOC 12% Zhou et al. (1996) Free Tropospheric air Simple box model based on 211± 104c Over predict CH4 chemistry, constrained to CH3COOH Liu et al. (1992) Free Tropospheric air Photochemical box model based 105±42 Over predict on CH4 Jacob et al. (1996) South Atlantic Photochemical box model 110 Over predict (Aircraft) includes CH4, inorganic chemistry and NMVOCs (exc. CH3OH+CH3CHO) Fried et al. (2003) Marine, Paciﬁc NASA Langley box model, con- (Average 2–4 km) Equivalent (Aircraft) strained by observations includ- 196±161 ing oVOCs Frost et al. (2002) Marine, North At- Photochemical box model in- (Average 0–2 km) Under predicted by a lantic, (Aircraft) cludes NMHC, and constrained 410±0.11 factor of 2 to measurements. However some values were taken from other campaigns.

formaldehyde concentration based on two models. The particularly the decrease in concentration in the afternoon “clean” model used only a very small subset of the MCM and evening of JD221. based on methane and CO chemistry whilst the “full-oxy” Figure 8b shows a comparison of the clean and full-oxy model was based on a mechanism originating from CO, models with the experimental data for the period 15–21 Au- CH , 23 hydrocarbons (including isoprene) (Lewis et al., 4 gust. The predictions from the full-oxy model are in reason- 2005), DMS, chloroform and three oxygenated compounds able agreement with the formaldehyde levels observed from (methanol, acetaldehyde and acetone) (Hopkins et al., 2003; UoL apparatus. In general the model does not do a particu- Lewis et al., 2005). During the simulations the model was larly good job in predicting the fine structure of the formalde- constrained to 15 min averages of the hydrocarbons, H , 2 hyde concentrations (an exception being the sharp increase in O , NO, NO ,H ), temperature and the measured photol- 3 2 2 concentrations in the early afternoon of 16 August, observed ysis rates of O , NO , HONO, HCHO, CH COCH and 3 2 3 3 by both experimental techniques), but this is to be expected CH CHO. The model included parameters for heterogeneous 3 from a box model. uptake and dry deposition and was run for several days (typically 2–3) to initialize the values of some species which were Figure 8b also highlights the significant contribution that not constrained. the higher hydrocarbons and particularly the oxygenated species make to the predicted formaldehyde concentration. Figures 8a and b show comparisons of the models with the This observation is in good agreement with the analysis by experimental data for two periods. Figure 8a is for JD221– Wagner et al. (2002), where even in the much less complex 222 during the westerly airflows when both instruments ob- conditions of the remote Indian Ocean, only 77% of HCHO served a pronounced diurnal variation. For these conditions originates from methane, the other 23% coming from ethane, the full-oxy model is clearly in better agreement with the ethene, propene, isoprene, acetone and DMS, all at concen- UEA data, although the shape of the diurnal pattern does trations significantly lower than those encountered at Mace not produce the diurnal profiles observed by either technique, Head. Our model highlights the importance of acetaldehyde www.atmos-chem-phys.net/6/2711/2006/ Atmos. Chem. Phys., 6, 2711–2726, 2006 2724 T. J. Still et al.: Ambient HCHO at NAMBLEX on formaldehyde generation, but unfortunately acetaldehyde westerly air masses (Figs. 5a, b) both experimental tech- was not considered by Wagner. niques show a decrease of approximately 150 pptv formalde- A rate of production analysis also emphasises the domina- hyde during this period. Assuming that physical processes tion (>70%) of the CH3O2+NO reaction as the loss mech- can be represented by a first order loss, this allows a predic- anism for CH3O2 for [NO] in the region of 10 pptv. Un- tion of the initial concentration at the start of the deposition der these conditions formation of the CH3OOH reservoir is only period. limited and therefore the potential for carbon removal via [HCHO] − 150 = [HCHO] e−kt (1) CH3OOH dry deposition is relatively small. 0 0 There have been several recent formaldehyde measure- For a sum of deposition and entrainment equal to ment/model comparisons in the North Atlantic environment 1×10−5 s−1, based on the reported deposition and entrain- (Weller et al., 2000; Frost et al., 2002; Fried et al., 2003) ment velocities of 0.4 cm s−1 (Wagner et al., 2002) and a in which all under predict measured formaldehyde concen- boundary layer height of 800 m (Norton et al., 2006), Eq. (1) trations to some degree. Weller et al. compared measured predicts [HCHO] of 600 pptv, in between the two obser- ± 0 average formaldehyde concentrations of 580 160 pptv from vations. Currently, uncertainties in boundary layer height a North Atlantic cruise with calculated values based on a box (±200 m) and deposition velocities (up to 100%) are such model. On the basis of measured alkane concentrations, they that the calculation range can encompass both observations. argued that methane should be the major source, but noted However, reductions in such uncertainties could help differ- that alkenes (not measured) could be a significant source. entiate between the two observations. The standard methane model under predicted the measurements by 250 pptv. Better agreement was found by introduc- 4.4 Implications for HOx production ing a formaldehyde production channel from Reaction (R2); Formaldehyde is an important photolytic source of radicals in the troposphere (Reaction R13). The MCM has been CH3O2 + HO2 → HCHO + H2O + O2 (R17) used to quantify the influence of the measured formaldehyde levels on [HO ]. Variations in formaldehyde levels based on arguments by Ayers et al. (1997). Making channel x have relatively little influence on OH radical concentrations (Reaction R17) 40% of the total reaction increased average as while increased formaldehyde generates more HO rad- formaldehyde concentrations by 80 pptv. Increasing the rate x icals, formaldehyde is also a significant HO sink via Reac- coefficient for OH abstraction and changing the branching tion (R10). Using either set of formaldehyde measurements ratios for the OH + methylhydrogenperoxide reaction within changed the predicted OH concentration by <2% under typ- values suggested in the literature increased formaldehyde ical westerly conditions. Increased formaldehyde concentra- by 50 pptv over the base model. Frost et al. generally un- tions have a more pronounced positive effect on HO chem- derestimated airborne formaldehyde measurements from the 2 istry, although the relationship is not 1:1, being buffered NARE 97 campaign by between 130 and 180 pptv (∼ factor by relatively complex radical interconversion. Increasing two) using a methane/hydrocarbon model. They carried out a formaldehyde concentrations by 50% produces between a sensitivity analysis and concluded that it was not possible to 15–25% increase in HO concentrations. It should be noted bridge the gap between measurement and model within the 2 that currently (Sommariva, 2004) the MCM tends to over- currently understood uncertainties in the model parameters predict HO concentrations by up to a factor of two. As noted but noted that VOCs not measured or considered in the model 2 earlier, formaldehyde may have a more pronounced influence could represent the missing source. Finally, Fried et al. report on HO production in other environments e.g. where a lack generally good agreement between measurement and model x of ozone or water prevents OH generation via O(1D) produc- for the TOPSE 2000 experiment, although again there is sig- tion. nificant under prediction at high latitudes over North Atlantic regions (Fried et al., 2003). Our modelling work emphasises the importance of oxy- 5 Conclusions genated species and particularly acetaldehyde and methanol in generating significant concentrations of formaldehyde. We Formaldehyde concentrations, measured by two different would therefore suggest that these species, measured val- techniques during the NAMBLEX campaign appear to be ues of which were not considered in the above modelling correlated, but exhibit a significant offset. Joint calibrations studies, as a potential source of missing formaldehyde and and subsequent intercomparison experiments have failed to recommend that concurrent measurements of all oxygenated identify the cause of the systematic errors present in one, or species are considered for future campaigns. both sets of apparatus. The laboratory intercomparisons at During the night (from 20:00–04:00 GMT) deposition and NPL in the presence of water vapour were inconclusive and entrainment are the only removal processes for formalde- further work in planned in this area in a new atmospheric hyde. From the averaged diurnal formaldehyde profiles for chamber in Leeds.

Atmos. Chem. Phys., 6, 2711–2726, 2006 www.atmos-chem-phys.net/6/2711/2006/ T. J. Still et al.: Ambient HCHO at NAMBLEX 2725

Formaldehyde continues to be a difficult molecule to mea- References sure with intercomparisons reporting varying levels of agreement. A recent comparison from the BERLIOZ campaign Ayers, G. P., Gillett, R. W., DeServes, C., and Cox, R. A.: (Grossmann et al., 2003) reported formaldehyde measure- Formaldehyde production in clean marine air, Geophys. Res. ments from a DOAS system to be 30% higher than those Lett., 24, 401–404, 1997. Cardenas, L. M., Brassington, D. J., Allan, B. J., Coe, H., Alicke, from a commercial Hantzsch system. Still more recently, B., Platt, U., Wilson, K. M., Plane, J. M. C., and Penkett, S. an intercomparison in Milan (Hak et al., 2005) gave good A.: Intercomparison of formaldehyde measurements in clean and agreement between FTIR, DOAS, Hantzsch and DNPH tech- polluted atmospheres., J. Atmos. Chem., 37, 53–80, 2000. niques with the slopes of the Hantzsch and DOAS regression Carslaw, N., Creasey, D. J., Heard, D. E., Jacobs, P. J., Lee, J. line not significantly differing from one. However, it should D., Lewis, A. C., McQuaid, J. B., Bauguitte, S., Pilling, M. be noted that formaldehyde concentrations during this inter- J., Penkett, S. A., Monks, P. S., and Salisbury, G.: Compar- comparison (2–13 ppbv) were significantly higher than those isons of model concentrations of OH, HO2 and RO2 with mea- encountered at Mace Head. More work is required to iden- surements during EASE 97, J. Geophys. Res., 107(D14), 4190, tify systematic errors in formaldehyde detection, especially doi:10.1029/2001JD001568, 2002. at low concentration levels. Fried, A., Crawford, J., Olsen, J., Walega, J., Potter, W., Wert, B., Jordan, C., Anderson, B., Shetter, R., Lefer, B., Blake, D. R., Comparison with previous formaldehyde measurements in Blake, N. J., Meinardi, S., Heikes, B., O’Sullivan, D., Snow, J., the North Atlantic suggests typical concentrations between Fuelberg, H., Kiley, C. M., Sandholm, S., Tan, D., Sachse, G. W., 300–800 pptv under clean background conditions. The UEA Singh, H. B., Faloona, I., Harward, C. N., and Carmichael, G. data lie at the low end of this range, the UoL data just above R.: Airbourne tuneable diode laser measurements of formalde- the upper limit. hyde during TRACE-P. Distributions and model comparisons., J. Geophys. Res., 108(D20), 8798, doi:10.1029/2003JD003451, Previous campaigns have shown a strong positive corre- 2003. lation between formaldehyde and acetaldehyde and with a Fried, A., Lee, Y.-N., Frost, G., Wert, B., Henry, B., Drummond, HCHO to CH3HCO ratio of approximately 2:1. A good cor- J. R., Hubler, G., and Jobson, T.: Airborne formaldehyde mea- relation has been observed between the UoL formaldehyde surements over the North Atlantic during the 1997 NARE cam- data and acetaldehyde measurements from the University of paign: Instrument comparisons and distributions., J. Geophys. York. The ratio of concentrations is in agreement with previ- Res., 107, 4039, doi:10.1029/2000JD260, 2002. ous work. Fried, A., Wang, Y., Cantrell, C., Wert, B., Walega, J., Ridley, B., Atlas, E., Shetter, R., Lefer, B., Coffey, M. T., Hannigan, J., Both instruments show a distinct diurnal profile during Blake, D. R., Blake, N. J., Meinardi, S., Talbot, R., Dibb, J., periods of consistent westerly airflow with an amplitude of Scheuer, E., Wingenter, O., Snow, J., Heikes, B., and Ehhalt, D.: 150–250 pptv. This is the expected behaviour of a photo- Tunable diode laser measurements of formaldehyde during the chemical intermediate, but such behaviour has not always TOPSE 2000 study: Distribution, trends and model comparisons, been observed in previous campaigns, either due to sampling J. Geophys. Res., 108(D4), 8365, doi:10.1029/2002JD002208, frequency or because of fluctuations in the concentrations of 2003. Frost, G. J., Fried, A., Lee, Y.-N., Wert, B., Henry, B., Drummond, the various air masses arriving at the receptor. This latter J. R., Evans, M. J., Fehsenfeld, F. C., Goldan, P. D., Holloway, factor is especially important in shipborne cruises. J. S., Hubler, G., Jakoubek, R., Jobson, B. T., Knapp, K., Kuster, Modelling studies from the latter part of the campaign W. C., Roberts, J., Rudolph, J., Ryerson, T. B., Stohl, A., Stroud, predict concentrations between the two experimental mea- C., Sueper, D. T., Trainer, M., and Williams, J.: Comparisons of surements, but more importantly, highlight the importance box model calculations and measurements of formaldehyde from the 1997 North Atlantic Regional Experiment, J. Geophys. Res., of oxygenates in formaldehyde production. The involvement 107(D8), 4060, doi:10.1029/2001JD000896, 2002. of oxygenated species in formaldehyde production may ex- Gilpin, T., Apel, E., Fried, A., Wert, B., Calvert, J. G., Zhang, G. plain some of the under predictions of simple methane only F., Dasgupta, P., Harder, J. W., Heikes, B., Hopkins, B., West- models. The concentrations of formaldehyde predicted by berg, H., Kleindienst, T., Lee, Y. N., Zhou, X. L., Lonneman, W., the MCM for an earlier part of the campaign are in better and Sewell, S.: Intercomparison of six ambient formaldehyde agreement with UEA data. measurement techniques, J. Geophys. Res., 102(D17), 21 161– 21 188, 1997. Grossmann, D., Moortgat, G. K., Kibler, M., Schlomski, S., Bachmann, K., Alicke, B., Geyer, A., Platt, U., Hammer, Acknowledgements. The authors would like to acknowledge the M.-U., Vogel, B., Mihelcic, D., Hofzumahaus, A., Holland, help of P. G. Quincey, N. Martin at NPL and NERC for funding F., and Volz-Thomas, A.: Hydrogen peroxide, organic perox- the NAMBLEX project. T. J. Still and S. Al-Haider thank NERC ides, carbonyl compounds and organic acids measured at Pab- and the Kuwaiti Government respectively for the funding of PhD stthum during BERLIOZ, J. Geophys. Res., 108(D4), 8250, studentships. doi:10.1029/2001JD001096, 2003. Hak, C., Pundt, I., Trick, S., Kern, C., Platt, U., Dommen, J., Or- Edited by: P. Monks donez, C., Prevot, A. S. H., Junkermann, W., Astorga-Llorens, www.atmos-chem-phys.net/6/2711/2006/ Atmos. Chem. Phys., 6, 2711–2726, 2006 2726 T. J. Still et al.: Ambient HCHO at NAMBLEX

C., Larson, B. R., Mellqvist, J., Strandberg, A., Yu, Y., Galle, B., Pope, F. D., Smith, C. A., Davis, P. R., Shallcross, D. E., Ashfold, Kleffman, J., Lorzer, J. C., Braathen, G. O., and Volkamer, R.: M. N. R., and Orr-Ewing, A. J.: Photochemistry of formalde- Intercomparison of four different in-situ techniques for ambient hyde under tropospheric conditions, Faraday Discuss., 130, 59– formaldehyde measurements in urban air, Atmos. Chem. Phys., 73, doi:10.1039/b419227c, 2005. 5, 2897–2945, 2005, Qian, H. B., Turton, D., Seakins, P. W., and Pilling, M. J.: A Laser http://www.atmos-chem-phys.net/5/2897/2005/. Flash Photolysis/IR Diode Laser Absorption Study of the Reac- Harris, G. W., Klemp, D., Zenker, T., and Burrows, J. P.: Tune- tion of Chlorine Atoms with Selected Alkanes, Int. J. Chemical able diode laser measurements of trace gases during the 1988 Kinetics, 34, 86–94, 2002. Polarstern cruise., J. Atmos. Chem., 15, 315–326, 1992. Ramacher, B., Rudolph, J., and Koppmann, R.: Hydrocarbon mea- Heard, D. E., Carpenter, L. J., Creasey, D. J., Hopkins, J. R., Lee, surements during tropsopheric ozone depletion events: Evidence J. D., Lewis, A. C., Pilling, M. J., Seakins, P. W., Carslaw, N., for halogen chemistry., J. Geophys. Res., 104, 3633–3653, 1999. and Emmerson, K. M.: High levels of the hydroxyl radical in the Saunders, S. M., Jenkin, M. E., Derwent, R. G., and Pilling, M. winter urban troposphere, Geophys. Res. Lett., 31(18), L18112, J.: Protocol for the development of the Master Chemical Mech- doi:10.1029/2004GL020544, 2004. anism, MCM v3: Tropospheric degradation of non-aromatic Heard, D. E., Read, K. A., Methven, J., et al.: The North At- VOC., Atmos. Chem. Phys., 3, 161–180, 2003, lantic Marine Boundary Layer Experiment (NAMBLEX), Mace http://www.atmos-chem-phys.net/3/161/2003/. Head, summer 2002, campaign overview, Atmos. Chem. Phys., Seakins, P. W., Orlando, J. J., and Tyndall, G. S.: The Rate Coeffi- 6 2241–2272, 2006. cients and Fraction of Vibrationally Excited HCl from the Reac- Hopkins, J. R., Lewis, A. C., and Read, K. A.: A two-column tions of Chlorine Atoms with Methanol, Ethanol, Acetaldehyde method for long-term monitoring of non-methane hydrocarbons and Formaldehyde., Phys. Chem. Chem. Phys., 6, 2224–2229, and oxygenated volatile organic compounds, J. Environ. Monit., 2004. 5, 8–13, 2003. Solberg, S., Dye, C., Schmidbauer, N., Herzog, A., and Gehrig, R.: Hopkins, J. R., Still, T., Al-Haider, S., Fisher, I. R., Lewis, A. Carbonyls and nonmethane hydrocarbons at rural European sites C., and Seakins, P. W.: A Simplified Apparatus for Ambient from the Mediterranean to the Arctic, J. Atmos. Chem., 25, 33– Formaldehyde Detection via GC-pHID, Atmos. Environ., 37, 66, 1996. 2557–2563, 2003. Sommariva, R.: Understanding Field Measurements through a Mas- Hunter, M. C., Bartle, K. D., Lewis, A. C., McQuaid, J. B., My- ter Chemical Mechanism, School of Chemistry, Leeds, Univer- ers, P., Seakins, P. W., and Van Tilburg, C.: The use of the He- sity of Leeds, 2004. lium Ionization Detector for Gas Chromatographic Monitoring Sommariva, R., Haggerstone, A. L., Carpenter, L. J., Carslaw, N., of Trace Atmospheric Components, J. High Res. Chrom., 21, Creasey, D. J., Heard, D. E., Lee, J. D., Lewis, A. C., Pilling, 75–78, 1998. M. J., and Zador, J.: OH and HO2 chemistry in clean marine air Hunter, M. C., Bartle, K. D., Seakins, P. W., and Lewis, A. during SOAPEX-2, Atmos. Chem. Phys., 4, 839–856, 2004, C.: Direct measurement of atmospheric formaldehyde using gas http://www.atmos-chem-phys.net/4/839/2004/. chromatography-pulsed discharge ionisation detection, Analyti- Tanner, R. L., Zielinska, B., Uberna, E., Harsfield, G., and Mc- cal Communications, 36, 101–104, 1999. Nichol, A. P.: Concentrations of carbonyl compounds and the Jacob, D. J., Heikes, B., Fan, S. M., Logan, J. A., Mauzerall, D. L., carbon isotopy of formaldehyde at a coastal site in Nova Scotia Bradshaw, J. D., Singh, H. B., Gregory, G. L., Talbot, R., Blake, during the NARE summer intensive, J. Geophys. Res., 101(D22), D. R., and Sachse, G. W.: Origin of ozone and NOx in the trop- 28 961–28 970, 1996. ical troposphere: A photochemical analysis of aircraft observa- Wagner, V., Schiller, C., and Fischer, H.: Formaldehyde measure- tions over the South Atlantic basin, J. Geophys. Res., 101(D19), ments in the marine boundary layer of the Indian Ocean during 24 235–24 250, 1996. the 1999 INDOEX cruise, J. Geophys. Res., 106(D22), 28 529– Lewis, A. C., Hopkins, J. R., Carpenter, L. J., Stanton, J. C., Read, 28 538, 2001. K. A., and Pilling, M. J.: Sources and sinks of acetone, methanol, Wagner, V., von Glasow, R., Fischer, H., and Crutzen, P. and acetaldehyde in North Atlantic air, Atmos. Chem. Phys., 5, J.: Are formaldehyde measurements in the marine bound- 1963–1974, 2005, ary layer suitable for testing the current understanding of http://www.atmos-chem-phys.net/5/1963/2005/. CH4 photooxidation?, J. Geophys. Res., 107(D3), 4029, Liu, S. C., Trainer, M., Carrol, M. A., Hubler, G., Montzka, S. A., doi:10.1029/2001JD000722, 2002. Norton, R. B., Ridley, B., Walega, J., Atlas, E., Heikes, B., Hue- Weller, R., Schrems, O., Boddenberg, A., Gab, S., and Gautrois, bert, B. J., and Warren, W.: A study of the photochemistry and M.: Meridional distribution of hydroperoxides and formalde- ozone budget during the Mauna Loa Observatory Photochem- hyde in the marine boundary layer of the Atlantic (48◦ N–35◦ S) istry Experiment, J. Geophys. Res., 97(D10), 10 463–10 471, measured during the Albatross campaign, J. Geophys. Res., 105, 1992. 14 401–14 412, 2000. Norton, E. G., Vaughan, G., Methven, J., Coe, H., Brooks, D., Gal- Zhou, X. L., Lee, Y.-N., Newman, L., Chen, X. H., and Mop- lagher, M., and Longley, I.: Boundary layer structure and de- per, K.: Tropospheric formaldehyde concentration at the Mauna coupling from synoptic scale flow during NAMBLEX, Atmos. Loa observatory during MLOPEX 2, J. Geophys. Res., 101(D9), Chem. Phys., 6, 433–445, 2006, 14 711–14 719, 1996. http://www.atmos-chem-phys.net/6/433/2006/.

Atmos. Chem. Phys., 6, 2711–2726, 2006 www.atmos-chem-phys.net/6/2711/2006/